Show HN: Skeletal Animation in Your Browser via Dual Quaternion Linear Blending | Hacker News Reader

Show HN: Skeletal Animation in Your Browser via Dual Quaternion Linear Blending

chinedufn.github.io

75 pointschinedufn21 comments

Open articleView on HN

Neat demo. I have a couple of suggestions for camera control:

1. Allow the use of the mouse for camera control.

2. Use up/down arrows for rotation similar to what you do with left/right and have separate keys control zoom.

Thanks!

Ah yes great suggestions. Hastily threw the controls in a couple months ago and they're sure due for an update

( PR's welcome :) )

Especially needed since you can build a nice mouse trackball using dual quaternion exponential :)

ha this calls for an obligatory "PR's welcome :)"

Super cool! Any chance that you'll add OCR skinning [1]? It looks really cool, and seems to get the best of both linear blend and dual quaternion skinning!

[1] https://www.disneyresearch.com/publication/skinning-with-opt...

Thanks!

So admittedly I only heard of DLB a few weeks ago, and OCR skinning 10 seconds ago :)

But this looks pretty damn cool. Thank you so much for the link, I'm definitely going to read this paper and consider it for a standalone skeletal animation library that I've started working on

Dual quaternion skinning seems to have failed to catch on. Support was removed from Unreal Engine for example. Why is this?

Ah good question.

So I was using linear blend skinning and had those horrid artifacts.

Google.com told me that dual quaternion skinning was the solution, so more or less blindly went with that.

santaclaus mentioned OCR in another comment, so I'll be reading up on that

Why not use the standard weighted matrix blending? What does the quaternion approach provide?

So the problem with blending matrices is that you are not guaranteed to get a rigid transform (you might introduce scaling, for example). You are also not guaranteed to get the shortest path between your start and end matrices.

So you'll often end up with weird stretching and other artifacts when you blend your matrices.

Dual quaternions are guaranteed to produce rigid transforms when blended. You're also guaranteed to get the shortest path when blending.

If you checkout an older commit in the repo such as - fe7cba263cc4f0ecc3ed96fb9288ed17469dd539 and `npm run demo` you'll see an example of the problem with matrix linear blending.

You can also read more about this in "Skinning with Dual Quaternions" - https://www.cs.utah.edu/~ladislav/kavan07skinning/kavan07ski...

Probably one of the best responses to that type of question. Explanation, reasoning for your decision along with a commit referenced to investigate further. Good stuff and great demo, thanks for sharing!

You're too kind! Thanks!

Do those blended matrix artifacts appear similar to artifacts produced when blending quaternions but not scaling the result to the unit circle?

Yup yup, both non-normalized dual and regular quaternions will give you artifacts similar to blended matrices.

Thank you for explaining, and for the reference!

Sure thing!

Standard linear blending suffers from "candy wrapper" pinching. Dual quaternion blending is better at volume preservation.

https://a.fsdn.com/con/app/proj/mayadqskinning/screenshots/1...

Ah good idea linking to an image. Great illustration of the problem :)

I'm just getting an error in chrome

> Uncaught TypeError: Cannot read property 'clearColor' of null

I'll check back again later though, this sounds cool.

uh oh :(

Well if it still doesn't work, feel totally free to try it locally:

```

git clone https://github.com/chinedufn/collada-dae-parser

cd collada-dae-parser

npm install

npm run demo

```

Smooth movement anyway.

Reply on news.ycombinator.com